Fable
The two proofreaders in the archive room
The city archives had a rule: every external reply had to be proofread before it could be stamped. In the early years there was only one old proofreader. He had to verify that the reply content matched the archived records, and also annotate after each paragraph which volume and page it was cited from. The workload was too heavy, so the old proofreader merged the two steps into one—he would casually flip to a relevant page and mark it, and as long as it looked plausible, it passed.
There were no problems in the first few months. The answers citizens found had citations, and the format was neat, so nothing could be faulted during inspections from above. Everyone felt that merging the processes had improved efficiency, and no one asked whether the page numbers in those citations were actually correct. The old proofreader also got used to this rhythm, and occasionally even he felt some number was odd, but the next document arrived.
The problem arose from a land ownership reply. A citizen took a certificate issued by the archives to handle a transfer, and the window found that the cited archive volume number did exist, but that volume recorded a different plot of land. The citizen made three round trips, and only after checking the original did they discover that the reply itself was wrong—during early transcription, the areas of two plots had been mixed up.
During the post-mortem, the archive only then discovered the cost of merging two processes into one. The old proofreader admitted that when verifying whether the content was genuine, he mainly checked whether the citations had volume numbers; as for whether the content the volume numbers pointed to matched the reply, he never turned to that page to verify. The citations looked complete, but in fact they were masking errors at the content level.
The archive later split the process into two roles: the first person only verifies whether the reply content matches the original archive text, without writing citations; the second person only annotates the source of the verified content. Each task has its own set of check standards and can fail and be corrected independently.
A spot check six months later confirmed the benefit of the split. Among a batch of replies, three had incorrect citation page numbers—but the first person had independently verified the content, and the replies themselves were correct. In the past, this would have been passed through as "already proofread," and only when citizens found the citation errors would there be a complaint. After the split, the error stayed at the source-citation layer and did not affect the credibility of the reply.
The archive later wrote this lesson into its training manual: verifying authenticity is verifying authenticity, and citing sources is citing sources. Splitting the two steps does not mean doubling the workload; it only means each layer's errors can be caught at that layer.